Real World Uses for Nagios APIs Janice Singh

Agenda This presentation describes the Nagios 4 APIs and how the NASA Advanced Supercomputing at Ames Research Center is employing them to upgrade its graphical status display (the HUD) and explain why it’s worth trying to use them yourselves. Janice S Singh –

The HUD: Visualization of the Center Status Janice S Singh –

Monitored Resources Pleiades –11,176-node SGI ICE supercluster –184,800 cores (plus 32,768 GPU cores) Frontend systems Hyperwall visualization cluster Tape Storage - pDMF cluster NFS servers for /home on computing systems Lustre scratch filesystems with multiple servers PBS (Portable Batch System) job scheduler Ref: Janice S Singh –

Nagios 4 Application Programming Interface No additional setup required Returns JSON output – multi-language support … Three kinds of APIs –Archive –Object –Status Run from the cgi-bin directory Each of the APIs have a help query –domain.com/nagios/cgi-bin/statusjson.cgi?query=help –Also gives help if there is an error in the query Janice S Singh –

JSON example bin/objectjson.cgi?query=hostgroup&hostgroup=tools "data": { "hostgroup": { "group_name": "tools", "alias": "Tools Group", "members": [ "lamsdb", "lamsweb", "lnxsrv107", "nasrunner", "remedy", "reports" ], "notes": "", "notes_url": "", "action_url": "" } } Janice S Singh –

Original Data Flow network firewall (The Enclave) nsca nrpe ssh Cluster Compute Node Remote Node Web Server nagios nsc a nagios.cmd datagg nagios2.cmd nagios nagios web interface HUD orange - pipe file green - text file purple - web site HUD buffer nrpe Dedicated Nagios Node Janice Singh - HUD format downtime.log

Nagios 4 Benefits Upgrading simplified configuration file –Frequent system configuration changes –Error prone –Time consuming Was one file: 17,835 lines; now 23 files: 9,121 lines Majority of the cleanup was using hostgroups APIs eliminate datagg configuration file Janice S Singh –

Modified Data Flow network firewall (The Enclave) nrdp nrpe ssh Cluster Compute Node Remote Node Web Server nagios nagpopd nagios nagios web interface HUD green - flat file purple - web site HUD buffer nrpe Dedicated Nagios Node Janice Singh -

Data Transfer with NRDP vs NSCA Only using one pipe allows use of nrdp Removing datagg layer allows using nagios as it was intended nrdp’s larger file transfer simplifies process –Previously had to split/reassemble –Kernel limit may cause split/reassemble No longer need to overload the perfdata Janice S Singh –

API Type - Archive Gives historical information based on var/archives –Availability –Alerts –Notifications Based on timestamps that you give it bin/archivejson.cgi?query=availability&availabilityob jecttype=hosts& hostname=pbspl233b&starttime= & endtime=-0 Janice S Singh –

API Type - Object Mirrors what your nagios configuration is Hosts Services Contacts Commands Dependencies etc. bin/objectjson.cgi?query=hostgroup&hostgroup=tool s Janice S Singh –

API Type - Status Gives the current state of nagios checks Host Service Comment Downtime bin/statusjson.cgi?query=hostlist&formatoptions=en umerate& hostgroup=tools Janice S Singh –

Status API Post Processing The API return codes are different than nagios nagpopd converts for HUD Janice S Singh – Status Code (From Nagios To Hud): Pending: 1 => 6 Ok: 2 => 0 Warning: 4 => 1 Unknown: 8 => 3 Critical: 16 => 2

API GUI Tool Tool to figure out the variables for the APIs Display builds the query –Dropdowns provide only relevant variables –Displays and executes the query –Displays the resulting JSON –Hovering over the input gives you help tips domain.com/nagios/jsonquery.html Janice S Singh –

API GUI Tool Screenshot Janice S Singh –

API GUI Tool Hover Example Janice S Singh –

NAS Use of APIs nagpopd –datagg replacement –API for object model –API for status Scheduled downtime handling Janice S Singh –

Using API for nagpopd Uses objectJSON: Get the structure directly from the API Eliminates separate HUD config file –Duplicate effort –Human errors –Inertia (resist making changes) HUD configuration put into nagios config HUD content uses custom variables Janice S Singh –

NAS Local Process (nagpopd) Prepares HUD interfacing file: Object Model –Loaded at startup from API queries –Perl, but could be any OO language –Can apply to other processing needs –Specific processing via Service subclassing Some objects created from custom variables –Some hosts form Domains –MultiServiceGroup for shared filesystem servers Janice S Singh –

Object Model NII System:: Main System:: Config System:: Encode System:: Log System:: Query System:: Service2Object Objects:: Domain Objects:: Host Objects:: HostGroup Objects:: MultiServiceGroup Objects:: Service Objects:: A_Service Objects:: B_Service Objects:: Z_Service …

API Queries Object JSON used on startup to create the layout: –objectjson.cgi?query=hostlist&details=true –objectjson.cgi?query=hostgrouplist&details=true –objectjson.cgi?query=servicelist&details=true –objectjson.cgi?query=servicegrouplist&details=true Status JSON queried in a loop to get latest data –statusjson.cgi?query=servicelist&details=true Janice S Singh –

Processing Status Information Generic Service object: –Default process ::setStatus (no changes) –Default output ::writeHUDb (reformat for HUD) –Other output methods easily added ::writeJSON (planned) ::writeHTML (later version) others: MySQL commands, etc Service Subclass overrides methods: –Handles service unique process or output –One array maps service name to object.pm Janice S Singh –

Scheduled Downtime Handling Old solution edited downtime.log When host is down, nagios stops checking it Used to sync with external program (schedule) … –Previous solution required shadow host pleiades – actual host could be down Pleiades – shadow never down –Now able to use APIs… Janice S Singh – Host_a host_a

External Program Use External program (command line interface) $ schedule all ALEX 10/06/ :00-10:25 10/06/2014 Raid Maintenance SUSAN 10/06/ :00-10:25 10/06/2014 RAID maintenance REMEDY 10/06/ :30-12:40 10/06/2014 Restart to resolve issue. $ query=downtimelist&formatoptions=enumerate& details=true Merges and updates nagios downtimelist … Janice S Singh –

Updating downtimelist Use nagios external command feature –SCHEDULE_HOST_DOWNTIME; ; ; ; ; ; ; ; –SCHEDULE_HOST_DOWNTIME;pioneer; ; ;1;0;7200;janice;just a test Documentation described in: mandlist.php mandlist.php Janice S Singh –

Hiccups Fixed by Nagios support Custom variables didn’t show up in JSON output Percent signs broke the JSON … sometimes fatally JSON output was limited to 8k Newlines didn’t show up in output Janice S Singh –

Hiccups We have one plugin that outputs so much data it can’t be passed on the command line, so nrdp breaks. –Kernel limitation –Will have to send in packets Having to have nsca and nrdp work at the same time Janice S Singh –

Future Plans AJAX-style updates to only update the part of the page that needs it Use the other information we get from the APIs –When a service is acknowledged –Use archive data to display alerts based on trends Janice S Singh –

Conclusion Using nagios 4 APIs has made our process much easier and will do more so in the future Simplified configurations Enabled object model Improved the flow Can communicate with external processes Good customer support Janice S Singh –

Questions? Janice S Singh –

Thank You Janice Singh